Time-Domain vs Frequency-Domain Vibration Analysis — PatSnap Eureka
Time-Domain vs. Frequency-Domain Vibration Analysis for Rotating Machinery Faults
Engineers designing condition monitoring systems for rotating machinery face a fundamental methodological choice: analyse vibration signals in the time domain using statistical features, or transform them into the frequency domain to identify characteristic fault signatures. Understanding when to use each approach — and how to combine them — is critical for reliable fault detection and diagnosis.
Two Complementary Windows into Machinery Health
Vibration analysis is the dominant technique for non-invasive condition monitoring of rotating machinery — including rolling element bearings, gearboxes, shafts, and rotors. The fundamental question for any condition monitoring system designer is which representation of the vibration signal carries the most diagnostic information for the fault type being targeted.
Time-domain analysis examines the raw vibration waveform as it evolves over time. Engineers extract scalar statistical features — such as root mean square (RMS) amplitude, kurtosis, crest factor, peak value, and skewness — directly from the digitised time-series. These features summarise the overall energy level and impulsiveness of the signal without requiring any transformation. According to IEEE standards for machinery condition monitoring, time-domain features remain the most computationally efficient approach for embedded real-time monitoring systems.
Frequency-domain analysis transforms the time-domain signal into a representation showing the amplitude (and phase) of each sinusoidal frequency component present. The PatSnap Analytics platform tracks thousands of patents in this space, reflecting the importance of Fast Fourier Transform (FFT)-based methods as the dominant frequency-domain tool. Because each rotating component generates vibration at mathematically predictable characteristic frequencies, the frequency spectrum allows engineers to localise which specific component — and which type of defect — is responsible for anomalous vibration.
Neither approach is universally superior. Time-domain methods excel at detecting the presence of a fault and monitoring its severity progression. Frequency-domain methods excel at identifying the fault's location and physical mechanism. Sophisticated condition monitoring systems, as catalogued in PatSnap's engineering solutions database, routinely combine both approaches within a single diagnostic pipeline.
Time-Domain Features: What Each Metric Reveals
Each time-domain statistical feature has a specific diagnostic role. Understanding their individual sensitivities allows engineers to select the right feature set for a given fault type and machinery configuration.
RMS (Root Mean Square)
RMS measures the overall energy level of the vibration signal. It is the square root of the mean of the squared signal values over a time window. RMS increases gradually as faults develop and overall vibration energy grows — making it an effective global health indicator and trending metric. It is relatively insensitive to early-stage localised defects that produce brief impulses without significantly raising overall energy.
Best for: imbalance, misalignment, loosenessKurtosis
Kurtosis measures the fourth statistical moment of the amplitude distribution — specifically, how heavy-tailed the distribution is relative to a Gaussian curve. A healthy bearing produces a kurtosis value near 3. A bearing with a localised spall or pit generates repetitive impact pulses that elevate kurtosis to 10 or higher. Kurtosis is highly sensitive to early-stage bearing defects but can decrease at advanced fault stages as the signal loses its impulsive character and becomes more broadband.
Best for: early-stage bearing defectsCrest Factor
Crest factor is the ratio of the peak amplitude to the RMS value. It quantifies how much higher the signal peaks are relative to its average energy. Like kurtosis, crest factor is sensitive to impulsive fault signatures in early fault stages. However, as a fault progresses and RMS rises (denominator increases), crest factor can paradoxically decrease even as the fault worsens — a well-documented limitation that motivates the use of multiple complementary features rather than any single indicator.
Best for: early-stage impulsive faultsSkewness
Skewness measures the asymmetry of the amplitude distribution around its mean. A perfectly symmetric signal has zero skewness. Certain fault types — particularly those generating asymmetric loading on bearing raceways or gear tooth impacts that differ between approach and recession — produce non-zero skewness values. Skewness is less commonly used as a primary indicator but adds diagnostic value when combined with kurtosis and RMS in a multi-feature condition monitoring vector.
Best for: asymmetric fault loadingFeature Sensitivity and Fault Frequency Mapping
Selecting the right analytical feature for a given fault type requires understanding the sensitivity profile of each method across bearing, gear, shaft, and imbalance fault categories.
Time-Domain Feature Sensitivity by Fault Type
Kurtosis and Crest Factor show highest sensitivity to bearing defects; RMS is most reliable for imbalance and gear faults.
Rolling Element Bearing Fault Frequency Identification
Each bearing fault type generates vibration at a mathematically predictable characteristic frequency, enabling precise fault localisation in the frequency spectrum.
FFT, Envelope Analysis, and Spectral Kurtosis Explained
Frequency-domain techniques transform vibration signals to reveal the specific frequencies at which energy is concentrated — enabling fault localisation that time-domain statistics alone cannot provide.
Fast Fourier Transform (FFT)
The FFT decomposes a time-domain vibration signal into its constituent sinusoidal frequency components, producing a frequency spectrum (also called a power spectral density or PSD plot). Each rotating component generates vibration at characteristic frequencies determined by its geometry and rotational speed. Identifying peaks at these frequencies in the FFT spectrum — and at their harmonics — allows engineers to determine which component is faulty. Gear mesh frequency, shaft rotational frequency, and bearing fault frequencies (BPFO, BPFI, BSF, FTF) are the primary diagnostic targets. NIST standards for vibration measurement underpin FFT-based diagnostic protocols in industrial settings.
Identifies: gear faults, shaft faults, imbalance, misalignmentEnvelope Analysis (Amplitude Demodulation)
Envelope analysis is a three-step process: (1) bandpass-filter the raw signal around a structural resonance frequency excited by bearing impacts; (2) extract the envelope (instantaneous amplitude) of the filtered signal using the Hilbert transform; (3) apply FFT to the envelope signal. The resulting envelope spectrum reveals the repetition rate of impact events, which corresponds directly to bearing fault frequencies. This technique is the industry standard for detecting early-stage rolling element bearing defects because it can isolate the fault signature even when it is buried under noise or other vibration sources. PatSnap customers in industrial maintenance routinely apply envelope analysis within their condition monitoring systems.
Best for: early-stage rolling element bearing defectsSpectral Kurtosis
Spectral kurtosis extends the kurtosis concept into the frequency domain by measuring the kurtosis of the signal's amplitude at each frequency bin over time. This identifies which frequency bands carry the most impulsive (non-Gaussian) energy — precisely the bands that contain bearing fault signatures. The primary application of spectral kurtosis is the automatic selection of the optimal bandpass filter for envelope analysis, replacing subjective manual filter selection with a data-driven approach. The result is improved detection sensitivity for early-stage faults and reduced dependence on expert knowledge. The PatSnap Analytics tool tracks growing patent activity in adaptive spectral kurtosis algorithms.
Best for: optimal filter selection for envelope analysisGear Mesh Frequency Analysis
For gearboxes, the primary diagnostic frequency is the gear mesh frequency (GMF), calculated as the product of the shaft rotational frequency and the number of gear teeth. Healthy gears produce a GMF peak with low-amplitude sidebands. As gear tooth wear, pitting, or cracking develops, the amplitude of GMF sidebands increases, and additional harmonics appear. The pattern of sideband spacing and amplitude modulation allows engineers to distinguish between distributed wear (uniform sideband increase) and localised tooth damage (sidebands spaced at shaft frequency). ISO standards for gearbox condition monitoring define acceptance criteria based on GMF sideband levels.
Best for: gear tooth wear, pitting, crackingTime-Domain vs. Frequency-Domain: Head-to-Head Comparison
Search vibration analysis method patents with PatSnap Eureka
Find prior art, track assignees, and benchmark your condition monitoring approach against 2B+ global patent records.
Beyond Basic FFT: Advanced Signal Processing for Rotating Machinery
Modern condition monitoring systems combine time-domain and frequency-domain methods within multi-stage diagnostic pipelines that address the limitations of each approach in isolation.
Cyclostationary Analysis
Rotating machinery signals are inherently cyclostationary — their statistical properties repeat with the shaft rotation period. Cyclostationary analysis exploits this periodicity to extract fault signatures that are invisible to both standard time-domain statistics and conventional FFT. The cyclic power spectrum and spectral correlation function reveal modulation patterns between carrier frequencies (structural resonances) and cyclic frequencies (fault repetition rates), providing superior diagnostic resolution for complex multi-component machinery.
Wavelet Transform Analysis
The wavelet transform provides time-frequency representation of vibration signals — simultaneously showing when and at what frequency fault-related energy occurs. Unlike the FFT, which assumes signal stationarity over the analysis window, the wavelet transform is well-suited to non-stationary signals produced by variable-speed machinery or intermittent faults. Continuous wavelet transform (CWT) and discrete wavelet transform (DWT) decompositions are widely used for bearing and gear fault detection in variable operating conditions, as tracked in PatSnap's analytics platform.
Time-Domain vs. Frequency-Domain Vibration Analysis — Key Questions Answered
Time-domain vibration analysis examines raw vibration signals as they vary over time. Engineers extract statistical features such as RMS (root mean square), kurtosis, crest factor, and peak amplitude directly from the time-series waveform. These metrics are especially sensitive to impulsive fault signatures — for example, a bearing defect generating a sharp spike in the signal. Time-domain methods are computationally simple and well-suited for real-time monitoring, but they can struggle to isolate specific fault frequencies when multiple components are vibrating simultaneously.
Frequency-domain analysis transforms a time-domain vibration signal — typically using the Fast Fourier Transform (FFT) — into a spectrum showing the amplitude of each frequency component present in the signal. Because each rotating component (bearing, gear, shaft) generates vibration at characteristic frequencies determined by its geometry and rotational speed, the frequency spectrum allows engineers to identify which component is faulty by matching spectral peaks to known fault frequencies such as BPFO, BPFI, BSF, and FTF for bearings, or gear mesh frequency for gearboxes.
Kurtosis is a statistical measure of the 'peakedness' or impulsiveness of a signal's amplitude distribution. A healthy bearing produces vibration that approximates a Gaussian distribution with a kurtosis value near 3. When a localised defect such as a spall or pit is present on a bearing raceway or rolling element, it generates repetitive impact pulses that dramatically increase the kurtosis value — often to 10 or higher. Kurtosis is therefore a sensitive early-warning indicator for bearing faults and is widely used as a time-domain condition monitoring feature.
Envelope analysis (also called amplitude demodulation) is a hybrid technique that bridges time-domain and frequency-domain analysis. The raw vibration signal is first bandpass-filtered around a resonance frequency excited by bearing impacts, then the envelope (instantaneous amplitude) of the filtered signal is extracted, and finally an FFT is applied to the envelope signal. The resulting envelope spectrum reveals the repetition rate of impact events — which corresponds directly to bearing fault frequencies — even when the fault signature is buried under noise or other vibration sources. Envelope analysis is the industry-standard method for detecting early-stage rolling element bearing defects.
Rolling element bearings generate vibration at four characteristic defect frequencies, each determined by the bearing geometry and shaft rotational speed. BPFO (Ball Pass Frequency Outer race) corresponds to defects on the outer raceway. BPFI (Ball Pass Frequency Inner race) corresponds to inner raceway defects. BSF (Ball Spin Frequency) corresponds to rolling element defects. FTF (Fundamental Train Frequency, also called cage frequency) corresponds to cage defects. These frequencies are calculated from the number of rolling elements, contact angle, ball diameter, and pitch diameter. Identifying peaks at these frequencies in the vibration spectrum — or their harmonics — is the standard diagnostic approach for bearing fault localisation.
Standard FFT analysis reveals the average frequency content of a signal over the analysis window but cannot distinguish between stationary sinusoidal components and transient impulsive components that happen to share the same frequency. Spectral kurtosis extends the kurtosis concept into the frequency domain: it measures the kurtosis of the signal's amplitude at each frequency bin, identifying which frequency bands carry the most impulsive (non-Gaussian) energy. This allows engineers to automatically select the optimal bandpass filter centre frequency and bandwidth for envelope analysis, removing the subjectivity of manual filter selection and improving detection sensitivity for early-stage faults.
Need to search patent literature on vibration analysis methods? PatSnap Eureka gives you instant access.
Search Vibration Analysis Patents on EurekaAccelerate Your Rotating Machinery Diagnostics R&D
Join 18,000+ innovators already using PatSnap Eureka to search patent landscapes, track technology trends, and benchmark condition monitoring innovations across vibration signal processing, bearing fault detection, and predictive maintenance.
References
- IEEE — Institute of Electrical and Electronics Engineers: Standards for Machinery Condition Monitoring and Vibration Measurement
- NIST — National Institute of Standards and Technology: Vibration Measurement and Signal Processing Standards
- ISO — International Organization for Standardization: ISO 13373 Condition Monitoring and Diagnostics of Machines — Vibration Condition Monitoring
- PatSnap Analytics — Patent Landscape Analysis for Condition Monitoring and Predictive Maintenance Technologies
- PatSnap Customer Success — Industrial Condition Monitoring and Predictive Maintenance Case Studies
- PatSnap — AI-Native Innovation Intelligence Platform for R&D and IP Teams
All technical descriptions on this page reflect established engineering principles in rotating machinery condition monitoring and vibration signal processing. Patent landscape data sourced from PatSnap's proprietary innovation intelligence platform covering 2B+ global patent records.
PatSnap Eureka searches patents and research to answer instantly.